Demographic, Psychological, and Social ... - Semantic Scholar

Sex Res Soc Policy (2010) 7:176–200 DOI 10.1007/s13178-010-0017-y

Demographic, Psychological, and Social Characteristics of Self-Identified Lesbian, Gay, and Bisexual Adults in a US Probability Sample Gregory M. Herek & Aaron T. Norton & Thomas J. Allen & Charles L. Sims

Published online: 3 March 2010 # The Author(s) 2010. This article is published with open access at Springerlink.com

Abstract Using data from a US national probability sample of self-identified lesbian, gay, and bisexual adults (N=662), this article reports population parameter estimates for a variety of demographic, psychological, and social variables. Special emphasis is given to information with relevance to public policy and law. Compared with the US adult population, respondents were younger, more highly educated, and less likely to be non-Hispanic White, but differences were observed between gender and sexual orientation groups on all of these variables. Overall, respondents tended to be politically liberal, not highly religious, and supportive of marriage equality for same-sex couples. Women were more likely than men to be in a committed relationship. Virtually all coupled gay men and lesbians had a same-sex partner, whereas the vast majority of coupled bisexuals were in a heterosexual relationship. Compared with bisexuals, gay men and lesbians reported stronger commitment to a sexualminority identity, greater community identification and involvement, and more extensive disclosure of their sexual orientation to others. Most respondents reported experiencing little or no choice about their sexual orientation. The importance of distinguishing among lesbians, gay men, bisexual women, and bisexual men in behavioral and social research is discussed.

G. M. Herek (*) : A. T. Norton : T. J. Allen : C. L. Sims Psychology Department, University of California, One Shields Avenue, Davis, CA 95616-8686, USA e-mail: [email protected] A. T. Norton e-mail: [email protected] T. J. Allen e-mail: [email protected]

Keywords Lesbians . Gay men . Bisexuals . Public policy . Sampling . Survey research . Committed relationships . Politics and religion . Identity, community, and disclosure

“Empirical studies using nonrepresentative samples of gay men and lesbians show that the vast majority of participants have been involved in a committed relationship at some point in their lives [and] that large proportions are currently involved in such a relationship....” (American Psychological Association 2007, pp. 14–15) “...[D]ata are not available to indicate the exact number of lesbian and gay parents in the United States....” (American Psychological Association 2007, p. 25) “Most or many gay men and lesbians experience little or no choice about their sexual orientation.” (American Psychological Association 2003, p. 8) These three passages, all excerpted from amicus briefs submitted jointly by the American Psychological Association (APA) and other professional organizations in court cases involving gay rights, illustrate some of the ways in which descriptive data about the lesbian, gay, and bisexual population are relevant to policy debates. In each instance, the APA and its co-amici summarized current knowledge about an aspect of the US gay, lesbian, and bisexual population that was relevant to a question being considered by the court—respectively, how many gay men and lesbians are involved in a committed relationship, how many are parents, and how many experience their sexual orientation as a choice. Yet, in each instance, the briefs could not provide definitive population estimates because relevant data were not available from nationally represen-

Sex Res Soc Policy (2010) 7:176–200

tative samples of self-identified gay, lesbian, and bisexual adults. The need for data describing the gay, lesbian, and bisexual population is not limited to legal proceedings. As Black et al. (2000) have noted, such data are relevant to a wide variety of policy debates in the USA, including those about initiatives designed to prohibit discrimination based on sexual orientation, public policy concerning the provision of benefits to same-sex couples, military policy concerning service by openly gay personnel, and lesbian and gay parental rights. They observed that “informed policy analysis about these issues requires accurate demographic information about the lesbian and gay population” (Black et al. 2000, p. 139). Population data describing lesbians, gay men, and bisexuals also have important scientific implications insofar as they can inform researchers who study the gay, lesbian, and bisexual population. Examination of demographic, social, and psychological patterns in the population, for example, can highlight gaps in current scientific knowledge and suggest hypotheses for empirical testing. Reliable estimates of the extent to which various characteristics and experiences are present in the sexual-minority population can also assist researchers in interpreting data from nonprobability samples and assessing their likely generalizability. To date, however, most social science knowledge about people who identify as gay, lesbian, or bisexual has been based on data from nonprobability samples. These samples have been recruited through such venues as clubs, cafes, and commercial establishments catering to gay men, lesbians, and bisexuals; neighborhood and community events; community-based organizations; local and national publications; e-mail lists and web-based communities; and friendship networks (e.g., Bell and Weinberg 1978; Bradford et al. 1994; Herek et al. 1999; Martin and Dean 1990; Riggle et al. 2005; Rothblum et al. 2004; Rothblum and Factor 2001). Researchers have also used public records to recruit specific groups, such as same-sex couples who have married or legally registered their partnership in states where they are allowed to do so (Balsam et al. 2008; Rothblum et al. 2008). Although the data collected from such samples are sources of important information, the extent to which their participants represent the larger population is unknown (Harry 1986; Meyer and Colten 1999; Sell and Petrulio 1996). It has often been assumed that traditional probability sampling methods—which permit assessment of sampling error and whose results can be generalized beyond a specific sample—are not feasible with lesbians, gay men, and bisexuals because nonheterosexuals constitute only a small proportion of the population and because sexual stigma deters some individuals from disclosing their homosexual or bisexual orientation to researchers. Con-

177

cerns about the limitations of findings from convenience samples, however, have fostered the development of innovative strategies for obtaining probability samples of gay, lesbian, and bisexual people (Cochran and Mays 2006; Meyer and Wilson 2009). For example, researchers have used various methods to identify nonheterosexuals in large national probability samples (Badgett 1995; Cochran and Mays 2006; Edelman 1993; Harry 1990; Laumann et al. 1994) and have applied probability sampling methods to specific settings or venues where sexual-minority individuals are known to be concentrated (Blair 1999; Diaz et al. 2004; Diaz et al. 1996; Stall and Wiley 1988). When examining this body of research, it is important to note that sexual orientation is a multifaceted construct that encompasses sexual attraction, sexual behavior, personal identity, romantic relationships, and community membership (Herek 2000; Sell 2007). Most social and behavioral research has operationally defined sexual orientation in terms of attraction, behavior, or identity, or some combination of these constructs. Which of these definitions is most appropriate for a particular study depends on the research goals (Sell and Silenzio 2006). For example, studies of sexually transmitted diseases among men who have sex with men might optimally focus on sexual behavior, whereas research on experiences stemming from one’s status as an openly gay, lesbian, or bisexual individual would, ideally, operationalize sexual orientation in terms of identity. However, even in studies for which sexual orientation identity is the relevant variable, researchers employing existing data sets based on large probability samples have often had to operationalize sexual orientation in terms of sexual behavior simply because most surveys have not collected data about identity. In many studies of economic discrimination that use national survey data sets, for example, the results have been characterized in terms of disparities between heterosexual workers and their gay or lesbian counterparts (e.g., Badgett 1995; Berg and Lien 2002; Blandford 2003). Although the terms “heterosexual,” “gay,” and “lesbian” suggest a focus on identity, limitations of the available data dictated that the operational definitions of sexual orientation be based on self-reported sexual behavior, from which the researchers inferred respondents’ sexual orientation identity. Although unavoidable, such use of sexual behavior as a proxy for identity and community membership is limiting for several reasons (see Herek et al. 2007). For example, it inevitably excludes gay, lesbian, and bisexual individuals who were not sexually active during the specified time period (e.g., Carpenter 2005). Moreover, the population of individuals who have experienced same-sex attractions or engaged in same-sex sexual behavior includes many people who do not identify as lesbian, gay, or bisexual (e.g., Cochran and Mays 2006; Laumann et al. 1994). Insofar as

178

much of the stigma directed at gay, lesbian, and bisexual people finds behavioral expression when others become aware of their sexual orientation identity (e.g., Herek 2009b), the experiences of self-identified gay, lesbian, and bisexual people are likely to differ in important respects from, say, self-identified heterosexuals with incidental same-sex attractions or sexual behavior. Some studies with probability samples have operationalized sexual orientation in terms of identity, but they have been limited by small sample sizes.1 For example, the National Health and Social Life Survey collected data about respondents’ sexual behavior, attractions, and sexual orientation identity. However, the sample ultimately included only 24 women who identified as lesbian or bisexual and only 39 men who identified as gay or bisexual (Laumann et al. 1994). Similarly, the National Survey of Midlife Development in the United States asked respondents to label their sexual orientation as heterosexual, homosexual, or bisexual. Of the approximately 3,000 respondents in this national probability sample, only 41 identified as homosexual and only 32 as bisexual (Mays and Cochran 2001). Such small numbers clearly preclude extensive analysis of self-identified lesbians, gay men, and bisexuals. Other studies using probability samples have obtained larger numbers of self-identified lesbian, gay, and bisexual respondents, but the samples have been restricted to specific US states (Carpenter 2005) or cities (Blair 1999; Sell et al. 2007) or to gay neighborhoods or venues in specific cities (Diaz et al. 1996; Stall and Wiley 1988). These studies have yielded invaluable data, but their findings may not be generalizable beyond those settings. Another important limitation is that the data from probability samples have generally not permitted separate analyses of self-identified lesbians, gay men, bisexual women, and bisexual men. As noted previously, some studies that directly assessed sexual orientation identity have yielded samples that were simply too small to permit separate analyses of subgroups (e.g., Laumann et al. 1994; Mays and Cochran 2001). In other studies, the sexual orientation question was not framed in a manner that permitted differentiation between bisexual and homosexual respondents. For example, exit polls conducted in conjunction with national elections have asked respondents to indicate 1

The problem of small sample size is not restricted to studies that have focused on sexual orientation identity. For example, an analysis of data from male respondents in the third National Health and Nutrition Examination Survey Studies (N=3,648) yielded a weighted total of 79 men who reported any same-sex sexual behavior during their lifetime (Cochran and Mays 2000). A 1985 ABC News– Washington Post poll recruited a national probability sample of men and included a question about sexual attraction. Of the 663 respondents, 16 reported that they were attracted to members of their same sex and another five volunteered that they were attracted to both men and women (Harry 1990).


whether they are gay, lesbian, or bisexual without differentiating among these groups (Edelman 1993; Hertzog 1996). Yet, empirical research with nonprobability samples suggests that important differences may exist among sexual-minority subgroups. For example, lesbians may differ from gay men in their likelihood of being involved in an intimate relationship (Peplau and Fingerhut 2007), bisexuals may differ from lesbians and gay men in the extent to which they are open about their sexual orientation and feel connected to a sexual-minority community (Balsam and Mohr 2007), and lesbians and bisexual women may differ from gay and bisexual men in the extent to which they manifest self-directed stigma (Balsam and Mohr 2007; Herek et al. 2009). Whether or not these findings can be generalized beyond the specific samples in which they were initially observed is as yet unknown, but they highlight the value of collecting data from probability samples that are sufficiently large to permit comparisons among gender and sexual orientation subgroups. This article uses data from a national probability sample of self-identified gay, lesbian, and bisexual adults to estimate population parameters on a variety of demographic, psychological, and social variables. Recognizing that sexual orientation subgroups may differ, we also compare and contrast gay men, lesbians, bisexual men, and bisexual women on each variable. Rather than testing specific hypotheses, our central goal is to report basic descriptive data about self-identified gay, lesbian, and bisexual adults. Although an overwhelming number of questions about potentially interesting and important characteristics of the sexual-minority population could be generated, practical considerations limited the number of variables that could be assessed. Guided mainly by our review of policy studies and amicus briefs from scientific and professional organizations that have addressed topics for which data about the US population of self-identified gay, lesbian, and bisexual adults would be relevant (e.g., American Psychological Association 1986, 2003, 2007; Belkin 2008; Black et al. 2000; Egan and Sherrill 2005; Herek 2006; Schaffner and Senic 2006), we focused on variables in four categories. First, we examined the basic demographic characteristics of this population, including age, educational background, and race and ethnicity. We also examined key variables identified by Black et al. (2000) as warranting description, including geographical distribution, household structure, and military veteran status. Second, consistent with the present study’s focus on adults who identify as gay, lesbian, or bisexual, we report descriptive data about key aspects of sexual orientation identity. These include the extent to which respondents used various identity labels in describing themselves; felt committed to their sexual orientation identity; had disclosed their sexual orientation to others; and were involved with


the gay, lesbian, and bisexual community. We also assessed the extent to which respondents perceived they had chosen their sexual orientation, an issue that has often been raised in policy debates and in legal discussions of gay, lesbian, and bisexual rights (see, for example, the 2003 APA amicus brief quoted at the beginning of this article; see also Herman 1997). Third, recognizing the importance of religious and political institutions in shaping contemporary policy and public opinion affecting gay, lesbian, and bisexual people, we assessed several aspects of respondents’ religious and political involvement. Although it is widely recognized that the condemnation of homosexuality that characterizes many religious denominations often creates conflicts and challenges for gay, lesbian, and bisexual people, there has been relatively little examination of the role that religion plays in the lives of sexual-minority individuals (Rodriguez and Ouellette 2000). We obtained descriptive data concerning respondents’ affiliation with a religious denomination, their participation in religious services, and the importance of religion in their daily lives. In the realm of political involvement, national exit poll data have suggested that lesbian, gay, and bisexual voters tend to be liberal and identify with the Democratic Party (e.g., Edelman 1993; Hertzog 1996). We assessed the extent to which these attributes characterize the larger lesbian, gay, and bisexual population. Finally, relevant to ongoing national debates about marriage equality and lesbian and gay parenting (e.g., Herek 2006), we collected data concerning respondents’ current relationship and parental status, as well as their future aspirations related to marrying. We also asked respondents about their general attitudes toward civil unions and marriage rights for same-sex couples.2

Method The study employed a probability sample of Englishspeaking, self-identified lesbian, gay, and bisexual adults residing in the USA. The sample was drawn from the Knowledge Networks (KN) panel, a large (approximately 40,000 households at the time of data collection) probability sample of English-speaking US residents who were recruited through random digit dialing (RDD) methods. Upon initially joining the KN panel, respondents agreed to participate regularly in on-line surveys and were provided with free Internet access and equipment if they did not already have it. Thus, in contrast to Internet studies with volunteer samples recruited via the Web, the KN panel includes individuals who would not otherwise have Internet access because of their 2 Data about hate crime victimization and related experiences among members of this sample are reported elsewhere (Herek 2009a).

179

financial or social situation. Reflecting this fact, KN samples more closely match the US population than do other Internet samples. Indeed, they are demographically similar to the RDD samples used in traditional telephone surveys (Chang and Krosnick 2009; see also Berrens et al. 2003) and have been used extensively in academic research (for examples, see Knowledge Networks 2009). Sample and Procedure All KN panel members routinely answer a battery of background questions, including one about their sexual orientation (“Are you yourself gay, lesbian, or bisexual?”). A probability sample of 902 English-speaking adults (≥18 years of age) was drawn from the subset of all panel members who had previously responded affirmatively to this question. Following standard KN procedures, they each received an e-mail invitation to complete the survey at their convenience. A follow-up e-mail was sent to nonresponders after approximately 1 week. Neither invitation mentioned sexual orientation. As with all KN surveys, panel members were free to decline to participate. A total of 775 individuals (86%) accessed the questionnaire between September 13 and October 7, 2005. In response to an initial screening question (described subsequently), six respondents declined to state their sexual orientation, and 50 indicated they were heterosexual.3 They were thanked for their assistance, and their survey was terminated. This 3 We hypothesized that these individuals were heterosexual respondents who had incorrectly characterized their sexual orientation on the original screening questionnaire (e.g., due to misunderstanding the question). However, we also recognized that some may have been gay, lesbian, or bisexual but reluctant to disclose this fact in the current questionnaire (e.g., out of concern that their responses might be seen by a household member who was unaware of their sexual orientation). We compared the personal characteristics of these respondents with those of the self-identified sexual-minority adults in the current sample. On most variables (including marital status, race and ethnicity, current employment status, residence in a metropolitan area, presence of children under 18 in their household, Internet access independent of KN, political party affiliation, and self-described political ideology), the 50 respondents who reported they were heterosexual differed from the self-identified sexual-minority sample. Although we cannot draw definitive conclusions, these patterns are consistent with the hypothesis that most of the 50 respondents were indeed heterosexual. Moreover, insofar as educational level is correlated with general questionnaire response validity (e.g., Krosnick 1991), the fact that these respondents had less formal education than others (42% had not attended college) is consistent with the hypothesis that many of them had misunderstood the original KN screening question. These analyses suggest that simply asking respondents whether they are “gay, lesbian, or bisexual”—with response options of “yes” and “no”—may not be an optimal strategy for ascertaining sexual orientation identity in national probability samples. The question on the current survey, which presented the different sexual orientations along a continuum and included the familiar term “straight” as a synonym for “heterosexual,” may have been easier to comprehend and answer accurately.

180

screening process left 719 self-identified lesbian, gay, and bisexual respondents who completed the questionnaire. Within that group, 56 households were represented by multiple respondents. In these cases, one respondent was randomly selected from the household for inclusion in the data set, yielding a final sample of 662. Taking into account all attrition in the KN panel since the earliest stage of RDD recruitment, the response rate for the present study was 30% (American Association for Public Opinion Research 2006 [Formula 3]). This rate is relatively high for contemporary commercial surveys (Holbrook et al. 2008). Measures The variables included in the questionnaire are described here, and the wording of most questions is reported in the tables. When appropriate, the question wording was tailored to respondents’ sexual orientation (bisexual vs homosexual) and gender. Basic Demographic Characteristics and Other Background Variables Information about respondents’ age, race and ethnicity, residence, location, and household composition had been routinely collected by Knowledge Networks in prior questionnaires. The present survey included a question asking whether the respondent was currently on active military duty, a member of the Military Reserves or National Guard, or a military veteran. Sexual Orientation Identity As noted previously, all respondents had reported they were gay, lesbian, or bisexual on a previously administered KN questionnaire. The present survey began with a screening question that asked respondents “Which of the following best describes your sexual orientation?” and provided five options arrayed on a continuum from homosexual to heterosexual. For male respondents, the options were (a) gay or homosexual; (b) bisexual, mostly attracted to men; (c) bisexual, equally attracted to men and women; (d) bisexual, mostly attracted to women; (e) heterosexual or straight. For females, the first response option was lesbian, gay, or homosexual, and options (b) and (d) were transposed. Respondents were asked how often they use various identity terms to describe themselves (“Gay,” “Lesbian” [women only], “Bisexual,” “Queer,” “Dyke” [women only], “Homosexual”). They were then asked to indicate their preferred term for characterizing their own sexual orientation (e.g., “Gay,” “Lesbian,” “Bisexual,” “Queer,” “Homosexual”). This label was subsequently inserted into questions that referred to the respondent’s sexual orientation or identity. This individualized item wording is indicated throughout the present article as [L/G/B/Q/H]. We used two measures to assess the strength of respondents’ commitment to their sexual orientation identity and to


the larger gay, lesbian, and bisexual community. First, three items assessing commitment to a sexual-minority identity were taken from the Internalized Homophobia Scale, or IHP (Herek et al. 1998; Herek et al. 2009): (1) “In general, I’m glad to be [L/G/B/Q/H]”; (2) “If someone offered me the chance to be completely heterosexual (‘straight’), I would accept the chance”; and (3) “I wish I weren’t [L/G/B/Q/H].” Second, two items assessing community identification were adapted from the Importance to Identity subscale of the Collective Self-Esteem scale (Herek and Glunt 1995; Luhtanen and Crocker 1992): (1) “My membership in the [L/G/B/Q/H] community is an important reflection of who I am” and (2) “Overall, my membership in the [L/G/B/Q/H] community has very little to do with how I feel about myself.” All of these items were presented with 5-point Likert-type response formats ranging from “strongly agree” to “strongly disagree,” with each respondent’s preferred identity label substituted for [L/G/B/Q/H]. Perceived choice about one’s sexual orientation was assessed with the question, “How much choice do you feel you had about being [L/G/B/Q/H]?” The response options were “no choice at all,” “a small amount of choice,” “a fair amount of choice,” and “a great deal of choice.” Respondents were asked their age when they first knew about their sexual orientation (“How old were you when you first knew or decided you were [gay/lesbian/bisexual]?”) and when they first disclosed it to another person (“How old were you the first time you told someone else that you are [gay/ lesbian/bisexual]?”). They were subsequently asked whether their mother or father knew about their sexual orientation and, if applicable, how many of their sisters and brothers knew about it. In addition, using a scale that ranged from 0 (not at all out to any of them) to 7 (completely out to all of them), respondents reported the extent to which they were “out of the closet (openly [L/G/B/Q/H])” to six additional groups: (1) “other relatives—not your immediate family,” (2) “your current heterosexual (‘straight’) friends,” (3) “your casual acquaintances who are heterosexual (‘straight’),” (4) “heterosexual (‘straight’) friends whom you knew before you came out,” (5) “your boss and other supervisors at work,” and (6) “the people you work with on a daily basis (other than your boss or supervisors).” A “doesn’t apply to me” response option was included for each group. Community involvement was assessed by asking respondents to “rate how important each of the following activities is to you these days. By important, we mean that you would feel differently about life and about yourself if you couldn’t do this activity.” The list of activities was adapted from a scale developed by Herek and Glunt (1995) and consisted of the following: (1) “Knowing what is going on in the local [L/G/B/Q/H] community,” (2) “Doing volunteer work in the [L/G/B/Q/H] community,” (3) “Giving money to [L/G/ B/Q/H] organizations,” (4) “Being politically active in the


[L/G/B/Q/H] community,” and (5) “Reading community newspapers and magazines for news about the [L/G/B/Q/H] community.” Each activity was rated on a 4-point scale (Not at all important, Somewhat important, Fairly important, Very important). Respondents also were asked whether they had ever engaged in a variety of activities related to lesbian, gay, or bisexual issues, including public expressions of opinion (“Wore a button, posted a sign, or displayed a bumper sticker”); participating in a rally, march, or demonstration; contacting a government official; and contributing money to a lesbian, gay, or bisexual organization or cause. For comparison purposes, this series of questions was followed by a parallel set of items that asked whether the respondent had participated in the same activities for “a non-gay issue or cause—that is, something not related mainly to gay men, lesbians, or bisexuals.” Political and Religious Involvement Information about respondents’ political party affiliation and ideology (liberal, moderate, conservative) had been previously collected by Knowledge Networks. For the present study, respondents were asked whether they had voted in the most recent (2004) presidential election and, if so, for which candidate. They were also asked for information about their religious denomination, frequency of attendance at religious services during the previous 12 months, how much guidance religion provides in their day-to-day living, and (for respondents who reported affiliation with a religious denomination and any attendance at religious services) the extent to which their congregation includes lesbian, gay, and bisexual members. Relationships, Marriage, and Family Respondents were asked their current relationship status, their legal marital status, and how many children they have (including adopted children and stepchildren). Respondents currently in a relationship (including those who were married) were asked the gender of their partner. Those who were in a relationship but not married were asked whether they were cohabiting and the likelihood they would marry their partner if their state were to allow same-sex marriages (this conditional clause was omitted for respondents in Massachusetts, the only state where marrying a same-sex partner was legal at the time of data collection). Those who were not currently in a relationship were asked whether they would like to marry someday. Respondents’ attitudes toward marriage rights for same-sex couples were assessed with three items. Using a 5-point Likert-type response format ranging from “strongly agree” to “strongly disagree,” they indicated the extent to which they agreed or disagreed with each of the following statements: (1) “The law should allow two people of the same sex to marry each other.” (2) “There is really no need to legalize same-sex marriage in the United States.” (3) “The U.S. public isn’t ready for a debate about gay marriage.” In

181

addition, respondents were asked whether they strongly supported, somewhat supported, somewhat opposed, or strongly opposed state laws to create civil unions. An accompanying note explained that “civil unions are not marriage, but give a same-sex couple some legal protection in their home state in areas such as inheritance, health insurance, and hospital visits.” Data Analysis Tables 1, 2, 3, 4, 5, 6, 7, and 8 report population parameter estimates with 95% confidence intervals (CIs). The CIs facilitate comparisons among the four gender and sexual orientation subgroups and are preferable to p values because they indicate whether group differences are statistically significant while also providing additional information about effect size (Cumming 2008; Wilkinson and Task Force on Statistical Inference 1999). As reported subsequently, the four subgroups differed significantly in age, race, and educational level. We conducted analyses to assess whether these demographic patterns might account for the group differences in the outcome variables reported in Tables 3, 4, 5, 6, 7, and 8. For each outcome variable, therefore, we conducted two linear regression analyses (for continuous and ordinal outcome variables) or two logistic regression analyses (for categorical outcome variables). In the first equation, sexual orientation (homosexual vs bisexual), gender, and their multiplicative interaction term were entered. In the second equation, age, educational level, and race (dichotomized as Black vs nonBlack) were added as statistical controls. Except where noted in the subsequent text, inclusion of the control variables did not alter the patterns of significant differences among subgroups shown in the tables. Weighting The KN panel’s original RDD design yielded a simple random sample with equal probability of selection for all US households with a landline telephone. However, the actual probability of selection for individual respondents was affected by multiple factors (e.g., differences in household size, number of telephone lines). Design weights were assigned to each case to adjust for unequal probability of selection (e.g., Kish 1965).4 Because the use of weighted 4 Design weights were computed to account for (a) variations in the number of adults and telephone lines in the household; (b) oversampling of Blacks and Hispanics, households with prior Internet access, and, early in the life of the KN panel, residents of California, New York, Florida, Texas, and Central regional states; (c) undersampling of telephone numbers for which matching addresses were unavailable and of households in areas without MSN-WebTV coverage; and (d) slight overrepresentation of Chicago and Los Angeles during KN’s early pilot testing.

182

data necessitates special analytic techniques to correct standard errors (Lee and Forthofer 2006), analyses were conducted using STATA and SPSS Complex Samples, which permit such correction.

Results The sample consisted of 311 women (152 lesbians, 159 bisexuals) and 351 men (241 gay men, 110 bisexuals). Applying design weights, the weighted sample was 34.8% gay male, 14.6% lesbian, 26.9% bisexual male, and 23.7% bisexual female (Table 1).5 Unless otherwise indicated, the weighted data are used hereafter. Representativeness Check One challenge associated with evaluating the representativeness of a lesbian, gay, and bisexual probability sample is the general lack of comparison data from the population of self-identified sexual minorities. Even though the US Census does not collect information about individuals’ sexual orientation, however, Census data are available for a subset of the sexual-minority population, namely, adults who report they are members of a cohabiting same-sex couple. Taking advantage of the fact that such individuals were able to identify themselves in the 2000 Census, we assessed the present sample’s representativeness by comparing its members who were cohabiting with a same-sex partner to their counterparts in the Census data. These comparisons are shown in Table 2, with the 2000 Census data corrected for misclassifications of some heterosexual couples due to miscodings of the partners’ gender (Black et al. 2007).6 Except for mean age, the two groups do not differ significantly, as indicated by the overlapping 95% CIs. These findings are consistent with the conclusion that, apart from being slightly older, the current sample was generally representative of self-identified lesbian, gay, and bisexual adults in the USA.


Age, Race, Ethnicity, and Education As shown in Table 1, the mean age of respondents was 39,7 approximately two thirds were non-Hispanic White, and roughly one third had earned a college degree. Significant differences were observed in these variables among the sexual orientation and gender groups. Gay men (M= 45 years) were significantly older than all other groups, and lesbians (M=40 years) were significantly older than bisexual women (M=32 years). Only 43% of bisexual men were non-Hispanic White, compared with more than 70% of other respondents (21% of bisexual men were Hispanic and 29% were non-Hispanic Black). More homosexuals than bisexuals had earned a bachelor’s degree: 46% of gay men and 41% of lesbians reported having a degree, compared with only 16% of bisexual men and 28% of bisexual women. According to Census data from approximately the same time period, the mean age of US adults (18 and older) was 45, about 75% were non-Hispanic White, and 24% had earned a college degree.8 Thus, the present sample was younger than the US adult population, was less likely to be non-Hispanic White, and had a higher level of formal education. However, these patterns were not uniform across subgroups within the sample. Gay men’s mean age was not significantly different from that of US adult men, whereas the other sexual orientation groups were significantly younger. Patterns of race and ethnicity among gay men and lesbians did not differ from the US population, but bisexual men were less likely to be non-Hispanic White, and bisexual women were less likely to be Hispanic or nonHispanic Black.9 Finally, whereas gay men and lesbians were significantly more likely than the US adult population to have earned a college degree, bisexual men and women did not differ significantly from the population in this regard.

7

5 Among bisexuals, 27% (40 men, 33 women) reported they were mainly attracted to people of their same sex, 39% (34 men, 71 women) were mainly attracted to the other sex, and 34% (36 men, 55 women) were attracted equally to both sexes. Because of the large margin of error associated with groups of such small size, these three categories were combined for the analyses presented subsequently. 6 We are grateful to Dr. Gary Gates (UCLA Williams Institute) for his kind assistance in this regard.

Approximately one third of the respondents (34%) were under 30, 33% were 30–44 years old, and 33% were 45 or older. Gay men were underrepresented in the 18–29 age category, compared with bisexual men and women; bisexual men were underrepresented in the 30–44 category, compared with gay men and lesbians; and bisexual women were underrepresented in the 45 and older category, compared with gay men and lesbians. However, because of the small number of respondents in some subcategories, these comparisons across sexual orientation subgroups must be considered tentative. 8 Comparisons were made with data from the US Census Bureau’s American Community Surveys 2000–2003, using the UC Berkeley SDA interface (http://sda.berkeley.edu/archive.htm). 9 These patterns describe respondents who identified with a single racial or ethnic group. Our data do not permit intensive analyses of respondents reporting mixed race ancestry.


183

Table 1 Demographic characteristics of sample Variable

Gay men

Lesbians

Bisexual men

Bisexual women

241

152

110

159

662

34.8 28.9–41.2

14.6 11.7–18.2

26.9 19.1–36.4

23.7 18.8–29.3

100

23–89 45.3 a

18–79 40.1 b

18–40 36.6 bc

18–76 31.8 c

18–89 39.0

43.0–47.5

37.7–42.6

32.0–41.1

29.3–34.3

37.1–40.9

Race/Ethnicity Non-Hispanic White CI

70.5% ab 59.6–79.5

74.4% a 62.6–83.5

43.0% b 25.8–62.1

77.5% a 65.9–86.0

65.4% 56.5–73.2

Non-Hispanic Black CI

14.0% 7.6–24.3

12.8% 6.1–24.9

28.6% 10.9–56.8

5.2% 2.4–10.9

15.6% 9.1–25.5

Hispanic CI

11.3% 6.1–20.0

10.5% 5.0–20.6

20.6% 7.2–46.6

6.2% 2.7–13.6

12.5% 7.4–20.2

Other, mixed race CI

4.2% 1.0–16.1

2.3% 0.8–6.6

7.8% 3.2–17.7

11.1% 4.8–23.5

6.5% 3.8–11.0

Education (highest level) Less than high school CI

5.6% 2.2–13.5

7.8% 3.1–18.4

8.0% 3.1–19.4

8.9% 3.6–20.2

7.3% 4.6–11.6

High school diploma CI

19.5% 12.6–29.1

17.5% 9.8–29.3

47.2% 27.1–68.3

26.8% 16.5–40.4

28.4% 20.7–37.6

Some college (